In [1]:
import requests as r
import re
from bs4 import BeautifulSoup
import pandas as pd
import numpy as np
import seaborn as sns
import time
import matplotlib.pyplot as plt
%matplotlib inline
sns.set()
import warnings
warnings.simplefilter('ignore') #ignore the warnings, not the errors
WHY¶
spoiler¶
Real why¶
Data¶
In [4]:
df=pd.read_csv('NUMeconomic_freedom_index2019_exploratory.csv')
df
Out[4]:
Blue: Developed countries
In [11]:
pd.read_html('https://fr.wikipedia.org/wiki/Pays_d%C3%A9velopp%C3%A9#cite_note-1')[0].iloc[:,0]
Out[11]:
In [12]:
developed="""United States,
Canada,
Singapore,
Hong Kong,
Japan,
South Corea,
Israel,
Taiwan,
Norway,
Switzerland,
Germany,
Denmark,
Netherlands,
Ireland,
Finland,
Sweden,
Liechtenstein,
United Kingdom,
Luxembourg,
France,
Belgium,
Austria,
Slovenia,
Italy,
Spain,
Czech Republic,
Greece,
Estonia,
Andorra,
Cyprus,
Malta,
Slovakia,
Portugal,
Australia,
New Zealand"""
In [13]:
developed_list=[i.strip().replace(',','') for i in developed.split('\n')]
In [14]:
df['Developed']=df['Country Name'].isin(developed_list)
In [15]:
df["Developed"]=np.where(df["Country Name"]=='Korea, South',1,df["Developed"])
In [16]:
df["Developed"] = df["Developed"].astype(int)
In [17]:
# no andorra
# thanks Tiago
In [18]:
df
Out[18]:
Insights¶
In [19]:
import plotly.express as px
fig = px.scatter(y=df['2019 Score'], x=df['GDP per Capita (PPP)'], color=df.Developed, hover_name=df.Country, labels={"x": "GDP per Capita (PPP), $", "y": "Economic Freedom Score"})
fig.update_layout(hovermode="x")
fig.show()
GDP per capita prediction¶
DecisionTree (GDP per capita)¶
weighted values
RandomForest (GDP per capita)¶
LinearRegression (GDP per capita)¶
GDP growth (5 years)¶
DecisionTree (GDP growth (5 years)¶
RandomForest (GDP growth (5 years)¶
Linear Regression (GDP growth (5 years)¶
Developed or not¶
Decision tree¶
- I joined 2 dataframes but look how striking is the difference from real values and columns from that index
RandomForest¶
After all that..¶
Models¶
Dropping columns¶
In [23]:
cols_to_drop=[]
cols_to_drop.append('CountryID')
cols_to_drop.append('WEBNAME')
cols_to_drop.append('Region')
cols_to_drop.append('World Rank')
cols_to_drop.append('Region Rank')
cols_to_drop.append('Country')
cols_to_drop.append('2019 Score') #Not sure drop it or not
cols_to_drop.append('Country Name')
In [24]:
df.fillna(df.mean(),inplace=True)
Models¶
In [25]:
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import AdaBoostClassifier, RandomForestClassifier
from sklearn.metrics import confusion_matrix, accuracy_score, recall_score, f1_score, precision_score
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC
In [26]:
X_train, X_test,y_train,y_test = train_test_split(df.drop('Developed',axis=1), \
df.Developed, \
test_size=1/3, \
random_state=42)
cool models¶
In [27]:
# decision tree
model1=DecisionTreeClassifier()
model1.fit(X_train, y_train)
y_pred1=model1.predict(X_test)
conf1=confusion_matrix(y_test, y_pred1)
acc1=accuracy_score(y_test, y_pred1)
rec1=recall_score(y_test, y_pred1)
pre1=precision_score(y_test, y_pred1)
f11=f1_score(y_test, y_pred1)
display(conf1)
print('Accuracy', acc1)
print('Recall', rec1)
print('Precision', pre1)
print('F1', f11)
In [78]:
# Random forest
model1=RandomForestClassifier()
model1.fit(X_train, y_train)
y_pred1=model1.predict(X_test)
conf1=confusion_matrix(y_test, y_pred1)
acc1=accuracy_score(y_test, y_pred1)
rec1=recall_score(y_test, y_pred1)
pre1=precision_score(y_test, y_pred1)
f11=f1_score(y_test, y_pred1)
display(conf1)
print('Accuracy', acc1)
print('Recall', rec1)
print('Precision', pre1)
print('F1', f11)
In [79]:
# Random forest balanced
model1=RandomForestClassifier(class_weight='balanced')
model1.fit(X_train, y_train)
y_pred1=model1.predict(X_test)
conf1=confusion_matrix(y_test, y_pred1)
acc1=accuracy_score(y_test, y_pred1)
rec1=recall_score(y_test, y_pred1)
pre1=precision_score(y_test, y_pred1)
f11=f1_score(y_test, y_pred1)
display(conf1)
print('Accuracy', acc1)
print('Recall', rec1)
print('Precision', pre1)
print('F1', f11)
In [80]:
# adaboost
model1=AdaBoostClassifier()
model1.fit(X_train, y_train)
y_pred1=model1.predict(X_test)
conf1=confusion_matrix(y_test, y_pred1)
acc1=accuracy_score(y_test, y_pred1)
rec1=recall_score(y_test, y_pred1)
pre1=precision_score(y_test, y_pred1)
f11=f1_score(y_test, y_pred1)
display(conf1)
print('Accuracy', acc1)
print('Recall', rec1)
print('Precision', pre1)
print('F1', f11)
In [27]:
from xgboost import XGBClassifier
from catboost import CatBoostlassifier
#dont work
not cool models¶
In [81]:
knn=KNeighborsClassifier(5)
knn.fit(X_train,y_train)
y_pred1=knn.predict(X_test)
display(confusion_matrix(y_pred1,y_test))
display(accuracy_score(y_pred1,y_test))
In [82]:
lr=LogisticRegression(max_iter=10000)
lr.fit(X_train,y_train)
y_pred2=lr.predict(X_test)
display(confusion_matrix(y_pred2,y_test))
display(accuracy_score(y_pred2,y_test))
In [83]:
sv=SVC()
sv.fit(X_train,y_train)
y_pred3=sv.predict(X_test)
display(confusion_matrix(y_pred3,y_test))
display(accuracy_score(y_pred3,y_test))
In [84]:
nb=GaussianNB()
nb.fit(X_train,y_train)
y_pred4=nb.predict(X_test)
display(confusion_matrix(y_pred4,y_test))
display(accuracy_score(y_pred4,y_test))
In [ ]:
#wrap
Learnings¶
- models are nice
- a level of development of a country is predictable to 95%
- we must all protect our property
if I was to start from scratch...¶
- copy everything from my old notebook not to call a friend at 2h30 in the night to find the dataset
In [1]:
import plotly
plotly.offline.init_notebook_mode()